As an interested party, having some technical experience in cloud streaming, and knowing just enough to be dangerous, what is necessary could be accomplished 2 ways:
1) rely on track (relative gain) metadata, which as already mentioned above is not always available. Even if available, it might be difficult to correlate this metadata into a useful, consistent metric. Streaming services, could if they wanted to, always make this metadata available and usable, but it would be expensive for them to do. IIRC, spotify had/has an option to "normalize" their tracks, not sure how they accomplish this.
2) Processing each track locally, and computing a relative gain for each (computing your own relative gain). This is rather compute intensive, and to do so you need access to the entire track.
Within the context of a Lumin device, they don't always have access to the entire track before it is played, as it is likely streamed. You could delay playback until you've had a chance to pre-scan the track, but this would prove an unacceptable delay as network speed, device storage and compute power come into play.
A potential relative gain could be calculated as a track is played the first time, and then cached and reused on subsequent playbacks, but it could be a rather confusing user experience (eg - a track plays at 60db the first time and 70db subsequently; if unaware it'd be confusing). It would require some storage on the device for the cache (which may or may not be available). It would also obviously require opting in to Leedh.
In short, this is a difficult proposition, and I don't think we'll be seeing this in Lumin devices.